A method and apparatus for encoding multi-bit code digital speech by subtracting adaptive dither, inserting buried channel bits, and filtering, and an encoding and decoding apparatus for this method.

摘要

For encoding sound received as a stream of multibit input sample, from a finite length sequence of input samples, an instantaneous audibility threshold characteristic is determined. Next, a shaped, dither determined, signal is subtracted from the input samples to produce processing samples. Subtracting a dither signal dynamically ignores processing sample bits below the threshold. Next, quantization by a variable number of bits below the threshold is performed, while retaining all sample bits above the threshold. The ignored bits are replaced by the dither signal as buried channel bits at an adjustable number of bits per sample. Therefore, upgraded samples have non-ignored bits and buried channel bits are obtained and outputted. The noise is shaped through shape-filtering a difference between the upgraded samples and the processing samples, which shape-filtering may amend a received white-noise-like signal towards an actual threshold-versus-frequency characteristic.